Accurate measurement of the effects of clinical intervention relies on our ability to separate those effects attributable to the intervention from those attributable to characteristics of the patient. The most important factor that varies among patients is the severity of their illness. The need for good measures of severity of illness is especiall strong in AIDS, where treatments are continually evolving, and there is little secure knowledge regarding which interventions, if any, are effective. Our research evaluating two models of organizing inpatient AIDS care has lead to the identification of a new prognostic staging method that could greatly simplify AIDS research. Nurses' assessment of hospitalized AIDS patients' functional status is a more accurate predictor of inhospital mortality than more complex and expensive clinical staging systems. This was an unanticipated finding, and one that requires more research before we would recommend its wider use. Th aim of this pilot study is to determine the reliability of the four- category global measure of functional status (with respect to activities of daily living) as evaluated by regular staff nurses.